Fatigue Bridge

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

EUROSTEEL 2017, September 13–15, 2017, Copenhagen, Denmark

Remaining fatigue life of steel railway bridges


Improved models for considering realistic behaviour

Paul Kugler*,a, Harald Unterwegera


a
Graz University of Technology, Institut of Steel Structures, Austria
kugler@tugraz.at, h.unterweger@tugraz.at

ABSTRACT
Many steel railway bridges will reach the end of their scheduled design service life within the next
years. One can also observe some bridges, considerably younger than their scheduled design service
life, with insufficient remaining fatigue life and rarely cracks in structural elements. Mostly because
of the early applications of the welding technology and sometimes insufficient consideration of
fatigue effects, e.g. neglecting of high local secondary stresses. It is a great challenge and very
expensive, if a couple of bridges have to be replaced in a short period. Therefore, it is important to
have models for realistic fatigue verification to extend the service life.
The focus of this paper is to show different approaches - both on the load and the resistance side - to
verify an extended service life of steel railway bridges. For example, the big potential of the use of
real, measured stress spectra is presented, instead of the common approach according to the
Eurocode. In addition, the application of the consequent Miner’s rule results in another significant
reduction of equivalent stress ranges for application of the fatigue strength curves of the Eurocode.
Furthermore, the simplified and conservative assumptions regarding the fatigue resistance in
conventional assessment procedures according to the Eurocode (nominal stress approach) are
compared with alternative concepts for typical welded joints. For example, the geometric (hot spot)
stress concept, the effective notch stress concept and fracture mechanics. The Eurocode already
allows the use of the geometric stress concept, but gives little guidance. The practical
implementation of these concepts is shown for a representative detail of a steel bridge, considering a
variety of geometries.
Fracture mechanics approaches are state of the art in mechanical engineering, but still relatively
uncommon in civil engineering. These concepts allow for the explicit consideration of real or
assumed “non-detected” flaws or cracks in the component, thereby eliminating the need of an exact
knowledge of the past load history, up to the considered defect size. It is shown that the fracture
mechanics concept is suitable for steel bridges to determine the remaining fatigue life of critical
details.
Finally, a multilevel framework is presented to give guidance for engineers to evaluate the service
life of a specific steel railway bridge.

Keywords: service life of railway bridges, real traffic stresses, fracture mechanics

1 INTRODUCTION
Many steel railway bridges will reach the end of their scheduled, calculated design service life
within the next years and the question is how the service life can be extended. This paper shows
important findings and results of a two-year research project [1] with the aim to modify the current
conservative assumptions in service life calculations. It shows the big potential of different
approaches - both on the load and the resistance side - to verify an extended service life of steel
railway bridges. Examples are shown on the basis of a sample bridge and a typical reference detail.
The following topics are covered to quantify the remaining fatigue life as realistic as possible:
1. A comparison between the real, measured traffic load spectra and the prediction of the
Eurocode.
© Ernst & Sohn Verlag für Architektur und technische Wissenschaften GmbH & Co. KG, Berlin · ce/papers 1 (2017), No. 2 & 3
https://doi.org/10.1002/cepa.286 wileyonlinelibrary.com/journal/cepa 2388
| 2389

2. The calculation method of a damage equivalent stress range Δse with appropriate consideration
of partially damaging stress ranges under the initial constant amplitude fatigue limit ΔsD.
3. Alternative concepts to the nominal stress approach to calculate the fatigue resistance, also for
cases with cracked details.
4. A multilevel framework to give guidance for engineers to evaluate the service life of a steel
railway bridge, that (i) is conform to the design concept of the Eurocode, (ii) offers different
opportunities regarding the former topics 1-3, (iii) allows the inclusion of individual
measurements, (iv) is available for details without or with detected or assumed fatigue cracks
and (v) is useable in practice.

2 MEASURED TRAFFIC LOAD EFFECTS VERSUS EUROCODE PREDICTIONS

2.1 General remarks


The question in [1] was, how the design traffic load effects according to the Eurocode EN 1991-2
[2] and EN 1993-2 [3] (for the application of the fatigue design trains, types 1-8 in annex D of [2],
as basis for the damage equivalent factor l) correspond to the real traffic load effects of a railway
bridge in Austria. It was not clear at the beginning if there are higher or lower traffic load effects in
the main Austrian railway network. A concept to measure the traffic load effects (time history of
strains over 4 weeks (28 days) in each measuring point) was developed and tested on an open deck
steel railway bridge from the 60s, with insufficient remaining fatigue life. Fig. 1 shows a photo of
the selected truss bridge with a view of the bridge deck and markings of all measuring sections. It
shall be noted that the presented reduced load effect for fatigue cannot be used directly for other
bridges. More bridges have to be observed in future investigations, to get generally valid data. But it
could be shown, that it is possible to get appropriate traffic load data of railway bridges with a
relatively cheap and easily applicable measuring concept. The following example shows that there
is a big potential to increase the service life of an individual bridge, if it is possible to verify lower
traffic load effects for the fatigue damage over the previous service life.

Fig. 1. Photo of the sample truss bridge and a view of the bridge deck with marked measuring sections at the bottom
chord (H1, H2, H3) and the vertical members (H4, H5) of the main girder, at the cross-girders (Q1-Q4 = Qges
and Q5-Q12) and at the longitudinal girders - stringers (L1-L4 = Lges and L5-L10)

2.2 Determination of realistic stress spectra


To verify lower real traffic stress ranges at the relevant cross sections of the individual members of
a bridge, the resulting stress histories of train crossings have to be recorded over a representative
period. This is relatively easy and feasible with the use of strain gauges. The measured strain
histories can be used to calculate stress spectra and equivalent stress ranges Δse for common fatigue
verification procedures. Fig. 2 shows all measured and subsequent standardised stress spectra for
the marked cross sections in Fig. 1. In addition, Fig. 2 shows an upper and a lower stress spectrum,
which are later used to show the influence of the spectrum shape on the calculated fatigue life.
Worth mentioning is the fact that for the studied bridge (Fig. 1) with curved rails (R = 750 m) the
centrifugal forces have a significant influence on the stress ranges, increasing the stress ranges in
the outer main girder and stringers of each track. Due to limitations of the train speed (v = 60 km/h)

© Ernst & Sohn Verlag für Architektur und technische Wissenschaften GmbH & Co. KG, Berlin ∙ CE/papers (2017)
2390 |

for the analysed bridge also the dynamic amplification factor including the centrifugal force effect
was measured, based on the crossing of two locomotives with v = 5 km/h and v = 60 km/h.

Δs/Δsmax

n/nmax

Fig. 2. Measured and standardised stress spectra with an upper and lower limit spectrum

2.3 Determination of the equivalent stress range Dse


There are different damage accumulation hypotheses in literature to get the equivalent stress range
Δse for a given stress spectra, however the common procedure for civil engineers is the use of the
linear Palmgren-Miner rule, with the fatigue strength curves of the Eurocode 1993-1-9 [4]. The
reduced damaging effect of stress ranges under the constant amplitude fatigue limit DsD at 5×106 is
considered with a flatter slope of the S-N curve (m* = 2m-1), until the cut-off limit DsL at 1×108
cycles. Eq. (1) shows the calculation of an equivalent stress range Dse with a certain length ne
according to this procedure. This method is based on the modified Miner’s rule, an engineer-like
approach from E. Haibach [5], to consider the decrease of the constant amplitude fatigue limit
DsD(D) with an increasing damage sum D. An alternative method to this engineer-like approach is
the consequent consideration of the decrease of DsD(D) over the service life, also mentioned in [5].
Fig. 3 shows an example of the so called consequent Miner’s rule. Only stress ranges Dsi higher as
the current constant amplitude fatigue limit DsD(D) have a damaging effect and count for the
calculation of the damage sum D.
damaging effect of spectrum during D2
damaging effect of spectrum during D3

D, time

Fig. 3. Example for application of the consequent Miner’s rule with indices s, z and i for application of Eq. (2)

Eq. (2) shows the application of the consequent Miner’s rule. This rule leads to a significant
increase of the calculated fatigue life, based on Eq. (1), in particular for stress spectra whose
maximum stress range Dsmax is just a bit higher than the initial constant amplitude fatigue limit
© Ernst & Sohn Verlag für Architektur und technische Wissenschaften GmbH & Co. KG, Berlin ∙ CE/papers (2017)
| 2391

DsD. The following section shows that Dsmax and DsD are often very close together at critical
details of steel bridges, so the consequent Miner’s rule is highly recommendable.

1 ææ ö ö
Ds e = m × ç ç å Ds im × ni ÷ + Ds 1D- m × å Ds 2j m -1 × n j ÷ (1)
ç Ds >Ds ÷
ne èè i D ø Ds D >Ds j >Ds L ø
1
Ds e = (2)
æ z d

e å ç d å ni × Ds i ÷
m n × D /
d =s è i =1 ø
where Δss is the first stress cycle Δsi > ΔsD

2.4 Comparison of load effects - measurements versus Eurocode predictions


The effects of the traffic loads were measured over a period of 28 days at different cross sections.
Based on the strain histories stress spectra and equivalent stress ranges Δse were calculated. The
simulation of the stress histories due to train crossings, at the same cross sections, was carried out
with a detailed 3D model of the bridge and the traffic mix of the fatigue train types of the Eurocode
(train types T1 - T8 in [2]), which are the basis of the damage equivalent factor l. Because of the
limitation of the train speed on the bridge (v = 60 km/h), the comparison of stress spectra was done
for the quasi-static situation (v ≈ 0 km/h). Therefore the measured stress cycles were reduced by the
also measured effect of dynamic amplification and centrifugal force. A detailed schedule of the
trains enables the calibration of the length of the calculated stress spectra in the sense of the
Eurocode, with an ident overall tonnage as main train parameter for fatigue design. The stress
spectra for the bottom chord (H3) and the stringer (L3) are shown in Fig. 4. As alternative, a
calibration with an ident number of trains was investigated. This leads to a higher tonnage of the
simulation, that means the real trains are shorter and overall lighter, than the fatigue train types of
the Eurocode.

simulation simulation
Δs [MPa]

measurement

measurement

a) N b) N

Fig. 4. Comparison of the measured and the simulated stress spectra (for v ≈ 0 km/h) at a) section H3 and b) section L3

The objective comparison of the measured and the simulated stress spectra is only possible with the
equivalent stress range ∆σe, which is listed in Table 1 for all relevant cross sections. Here ∆σe is
based on the number of trains (one train crossing gives one cycle ∆σe). The columns Dse,2 to Dse,4
show also the result of the reduced damaging effect under the initial constant amplitude fatigue
limit DsD, via different Miner’s rules. In column ∆σe,1 no fatigue limit is considered, all stress
ranges Dsi have the same damaging effect, according to the elementary Miner’s rule. Column ∆σe,2
contains the modified Miner’s rule, according to E. Haibach [5], which is in fact the method of the
Eurocode without the cut-off limit DsL (Eq. (1) with ∆σL = 0). Column ∆σe,3 represents the
Eurocode approach (Eq. (1)) and column ∆σe,4 shows the results of the consequent Miner’s rule (Eq.
(2)). The value 0 in column ∆σe,4 means, that there is no damage at all, because the maximum value
of the spectrum Dsmax is lower than the constant amplitude fatigue limit DsD.

© Ernst & Sohn Verlag für Architektur und technische Wissenschaften GmbH & Co. KG, Berlin ∙ CE/papers (2017)
2392 |

To show the significant nonlinear effect of the value of the initial constant amplitude fatigue limit
DsD, in comparison with the maximum stress range Dsmax, there is a second evaluation of Dse,i in
Table 1 with a lower classification of the structural details. The first evaluation with an appropriate
classification (100 % FAT class) and the second with a 20 % reduction of the FAT class (80 % FAT
class). All cross sections show a surprisingly high reduction of the equivalent stress range Δσe of
the measurement, in comparison to the numeric simulation, which is also expressed in percentage
terms in Table 1. The better the reduced damaging effect under the initial constant amplitude fatigue
limit ΔσD is captured, the bigger is the reduction of Δσe. According to the Eurocode approach -
column Δσe,3 - for every cross section except H3, an unexpected high reduction of at least -50 %
was determined. This means an extension of the service life by a factor of (2/1)³ = 8. All results of
Δσe,i in Table 1 underestimate the fatigue effect, due to the missing effect of dynamic amplification
and centrifugal force.
Table 1. Δσe,i for stress spectra with ident overall tonnage; ne = 5125 trains in 28 days
100 % FAT class 80 % FAT class
cross
spectrum Δsmax ΔsD Δse,1 Δse,2 Δse,3 Δse,4 ΔsD Δse,1 Δse,2 Δse,3 Δse,4
section
simulation 54 66.3 27.2 20.4 19.9 0 53.1 27.2 23.5 23.3 22.4
H2 measurement 53 66.3 20.7 12.0 9.7 0 53.1 20.7 14.0 12.6 0.0
Δ% -24% -41% -51% - -24% -41% -46% -
simulation 54 66.3 24.5 17.1 16.1 0 53.1 24.5 19.7 19.2 17.6
H3 measurement 57.1 66.3 21.9 13.4 11.2 0 53.1 21.9 15.5 14.3 9.7
Δ% -11% -22% -30% - -11% -21% -26% -45%
simulation 40 52.3 35.4 24.0 23.1 0 41.8 35.4 27.9 27.1 0.0
H4 measurement 41.9 52.3 25.0 13.9 9.9 0 41.8 25.0 16.2 13.8 4.0
Δ% -29% -42% -57% - -29% -42% -49% -
simulation 39 52.3 33.5 22.3 21.4 0 41.8 33.5 25.9 25.1 0.0
H5 measurement 40.6 52.3 21.6 11.3 3.0 0 41.8 21.6 13.1 10.7 0.0
Δ% -36% -49% -86% - -36% -49% -57% -
simulation 46 58.9 50.4 36.7 34.9 0 47.2 50.4 42.5 41.8 0.0
L3 =
measurement 46.5 58.9 35.4 20.9 16.5 0 47.2 35.4 24.3 22.0 0.0
Lges
Δ% -30% -43% -53% - -30% -43% -47% -
simulation 48 58.9 42.2 29.2 27.3 0 47.2 42.2 33.9 32.7 26.0
Q3 =
measurement 55 58.9 31.7 18.3 14.0 0 47.2 31.7 21.2 18.7 11.2
Qges
Δ% -25% -37% -49% - -25% -37% -43% -57%

2.5 Further findings and interesting effects


· A complete implementation of the allowed distribution of axle loads by the rails on the
stringers, according to EN 1991-2 [2], shows only minor impacts on the stress results.
· An investigation of the beneficial influence of the rail (if activated or not), reducing the
stresses in the stringers, shows, that the rail is only partially activated in reality. It cannot be
used with its full stiffness of the cross section in the calculation to assist the stringer. A
conservative approach is to deny the contribution of the rail.
· The curvature of the rails (R = 750 m) on the bridge shows an unexpectedly high influence of
the centrifugal force on the stresses due to vertical loading (simulation without speed limitation
of the trains). The stress increase in comparison with the static vertical loading (without any
dynamic effects) was +20 % for the bottom chord of the outer truss girder, +10 % for the
vertical member of the outer truss girder and +40 % for the bottom flange of the outer stringer.
This effect is often neglected in practice and was further investigated at the institute [6].
Additional measurements on bridges with relevant rail curvature and supplementary numeric
calculations would be important to review and may correct (reduce) the height of the horizontal
load application for centrifugal forces (assumption at the moment: e = 1.8 m over the top edge
of the rail).

© Ernst & Sohn Verlag für Architektur und technische Wissenschaften GmbH & Co. KG, Berlin ∙ CE/papers (2017)
| 2393

3 ALTERNATIVE CONCEPTS TO CALCULATE THE FATIGUE RESISTANCE

3.1 General remarks


It is possible to use the geometric (hot spot) stress concept, or the effective notch stress concept, to
calculate the fatigue resistance of critical details. These alternative concepts and their theoretical
backgrounds are described in [5, 7-9]. The idea behind these concepts is, to model the geometric
discontinuities with finite elements, to be able to consider the stress concentrations in more detail.
This can avoid a potentially conservative detail classification of the nominal stress concept. In this
respect it is noted that there is only the geometric stress concept mentioned in the Eurocode [4] and
there is no further information, beside a detail category catalogue, given. This paper summarises the
findings from [1] for a practical implementation of these concepts on the basis of a reference detail.
Fig. 5a shows a welded stiffener connection at the bottom flange of a beam or stringer, a common
detail for many steel constructions with a detail category (FAT class) of 80 in the nominal stress
concept. For both concepts, the use of local submodels (as shown in Fig. 5 and Fig. 6) is very
helpful to decrease the calculation effort due to the required fine element mesh near the weld. It
should be noted that the local submodel has to be big enough to ensure correct local stresses, that
means that its boundaries are not influenced by the local stress concentrations.
3.2 Geometric (hot spot) stress concept
Geometric stresses are the stresses at the critical point on the detail surface (the hot spot, e.g. the
weld toe), without the nonlinear stress peak due to a local notch effect. There are a couple of
different variations of possible finite element meshes and related extrapolation formula given in [9],
to calculate the geometric stresses. A study of all these meshing and extrapolation variations in [1]
showed nearly similar geometric stress results. There was also no difference in the calculated
geometric stress between a double fillet weld (with a slit between the stiffener and the flange) and a
through welded connection. An important point is to use elements with reduced integration, both for
linear and quadratic elements. Fig. 5b shows the local model of a stringer with a stiffener in the
middle. The nominal stress at the top surface of the bottom flange is Δsbf = 80 MPa in this example.
Fig. 5c shows a detail of the local submodel for the calculation of the geometric stress. The stress
concentration in front of the transverse stiffener is clearly visible due to the limited colour scale
from 75 (blue) to 85 MPa (red).

ΔMy
L=6m

ΔMy

Δσbf = 80 MPa
a) b) c)

Fig. 5. a) Picture of a welded stiffener connection at the bottom flange of a stringer, b) local model of a stringer with a
stiffener, colour scale -100 (blue) to +100 MPa (red), c) local geometric stress submodel, colour scale 75 (blue)
to 85 MPa (red)

3.3 Effective notch stress concept


To consider the effect of the notch at the weld toe, the effective notch stress concept demands
elastic calculated local stresses. For that purpose, the real notch radius (r <<) is replaced by an
effective notch (r = 1 mm) in the finite element model [7]. According to the specifications in [9],
very small elements with an absolute size £ 0.25 mm are necessary. Fig. 6 shows the details of the
model to calculate the effective notch stress in front of the transverse stiffener with two local
submodels. The possible number of cycles can then be calculated with the resulting maximum
principal stress at the weld toe and the related S-N curve (m = 3, FAT = 225) [9].

© Ernst & Sohn Verlag für Architektur und technische Wissenschaften GmbH & Co. KG, Berlin ∙ CE/papers (2017)
2394 |

a) b) c)

Fig. 6. a) local model of a stringer with a stiffener, colour scale -100 (blue) to +100 MPa (red), b) local submodel,
colour scale: 0 (blue) to 110 MPa (red), c) local submodel to calculate the effective notch stress, colour scale: 0
(blue) to 176 MPa (red)

3.4 Comparison of the alternative stress concepts


Table 2 shows a fatigue life comparison of the geometric- and the effective notch stress concept
with the nominal stress concept, for different geometric variations. All geometric variations V1-V7
are classified as FAT 80 (for dimension w £ 50 mm) in the nominal stress concept. Fig. 7 shows the
nearly linear increase of the geometric stress sgeom over the attachment width w for all geometric
variations V1-V7 (snom = 80 MPa, detail classified as FAT 100 in the geometric stress concept).
The variations V3 and V6, with the unusually thick stiffener, are in this sense a limit value
consideration in this comparison. The parameter study shows for all common dimensions a higher
fatigue life, than for the nominal stress approach. Although it is interesting that the geometric stress
concept leads to a higher fatigue life for variations with tbf = 24 mm in comparison to variations
with tbf = 12 mm, because the effective notch stress concept has an opposing trend.
Table 2. Parameter study with a comparison of the stress concepts in terms of the fatigue life (N = number of cycles)
N geom N notch
tbf [mm] tst [mm] a [mm] w [mm]
N nom N nom
tst V1 12 8 5 18 160.0 % 210.7 %
V2 12 16 5 26 140.0 % 182.7 %
w V3 12 40 5 50 96.8 % 113.7 %
V4 24 8 5 18 174.9 % 152.9 %
V5 24 16 5 26 161.1 % 123.9 %
a tbf V6 24 40 5 50 122.7 % 84.4 %
V7 24 16 10 36 139.5 % -

Fig. 7. Geometric stresses sgeom for V1-V7 over the attachment width w (snom = 80 MPa)

4 FRACTURE MECHANICS

4.1 General remarks


With the S-N curve concept it is possible to determine the fatigue life of uncracked components.
But it is always necessary to consider the previous damage, due to fatigue loading in the past. The
advantage of fracture mechanics is, that it is not required to know the load history of a detail in the
past, to calculate a remaining fatigue life. The application of fracture mechanics is also possible, if
no cracks are found with a non-destructive testing method. Then a possible flaw or crack just under
the detection limit of the testing method has to be assumed. The theoretical background and the
required calculation tools are described in [10-12]. This section shows how fracture mechanics can
be used to determine the fatigue life of a steel bridge, with an example of a half-elliptical surface
© Ernst & Sohn Verlag für Architektur und technische Wissenschaften GmbH & Co. KG, Berlin ∙ CE/papers (2017)
| 2395

crack in front of a transverse stiffener, as shown in Fig. 8a. The example also illustrates the
significant influence of correct material parameters and the stress spectrum on the calculated fatigue
life. The statements and recommendations are based on an extensive crack propagation study in [1],
with different crack models and more than 1200 simulated parameter combinations.
4.2 General approach
At the beginning, it is important to define a crack model - like in Fig. 8a, a plate with a half-
elliptical surface crack under tension - and the initial crack size. Many different crack models are
found in [12], with the related formulae to determine the stress intensity factor K according to the
linear-elastic fracture mechanics approach (LEFM). In front of a transverse stiffener the assumption
of a half-elliptical surface crack, with an initial geometry of a0/c0 = 0.15, is recommended (see [13],
as background of [14]). Simple plate-shaped crack models are commonly used for many fracture
mechanics calculations and can lead to reasonably good results for many cases. The stress
concentration effect due to the transverse stiffener can be considered with local stresses Δsloc of a
finite element model, as boundary condition of the crack model (Fig. 5c). The calculated fatigue life
lasts from an initial crack size a0 to a critical crack size acrit. The latter can be calculated with the
failure assessment diagram (FAD) concept, that is described in [12]. It is recommended to use a
reduced combination of actions for low temperatures, as for the design verification to avoid brittle
fracture, according to EN 1993-1-10 [14]. In addition, residual stresses of about 100 MPa
(background of [14], if not known in more detail) have to be considered for welded details and the
calculation of the critical crack size. Fig. 8b shows the critical crack size for a half-elliptical surface
crack, as a function of the fracture toughness Kmat and a static load level of scrit = 0.7 fy (model
dimensions: 2w = 200 mm, t = 28 mm, a0 = 0.5ln(t) = 1.67 mm [13], c0 = a/0.15 = 11.11 mm,
formula valid for a £ 0.8 t). The crack propagation can be calculated with the integration of a crack
propagation law (e.g. the Paris-Erdogan law). In this example, the shape of the half-ellipse a/c was
not hold constant, it was controlled by the stress intensity factors Ka and Kc (see Fig. 8a, b).

90
80 cal
ccrit 70 uss
Δsloc Δsloc
60 Kmat = 140 MPa√m
50
acrit
Δsloc 40
Δs[MPa]

30

20 Kmat = 25 MPa√m
t constant amplitude loading
2w
upper stress spectrum
Ka 10
Kc 104 105 106 107 108 109 1010
a

2c N

a) b) c)

Fig. 8. a) crack model for a stiffener welded to the flange, a plate with a half-elliptical surface crack under tension, b)
critical crack size over the fracture toughness (start geometry a0/c0 = 0.15), c) fatigue life diagram for different
parameter combinations

The fatigue life for this crack configuration is shown in Fig. 8c, for a variety of parameter
combinations. The values of the fracture toughness range from Kmat = 25 to 140 MPa√m, with a
step size of 5 MPa√m. The continuous lines stand for a constant amplitude loading, the dashed lines
for an upper stress spectrum, as conservative consideration of all measured stress spectrum shapes
(see Fig. 2). The vertical axis shows the value of the constant amplitude loading Δs or the
© Ernst & Sohn Verlag für Architektur und technische Wissenschaften GmbH & Co. KG, Berlin ∙ CE/papers (2017)
2396 |

maximum stress Δsmax of the upper stress spectrum. The significant influence of the fracture
toughness and the shape of the stress spectrum is obvious.
4.3 Material parameters
For a simple crack propagation calculation at least the fracture toughness Kmat = KIc, the crack
propagation rate da/dN, the threshold stress intensity factor range DKth (below which crack
propagation can be considered as negligible) and the parameters of the Paris-Erdogan law C and m
are required. The values of fracture mechanics material parameters are subject to great variation. It
is recommended to carry out tests in specialized laboratories, to determine the needed material
parameters of every investigated structure, because they have a large impact on the calculated
fatigue life. Otherwise, they have to be conservatively defined with the help of literature sources,
like [11, 12]. In [11] the lower limit value of Kmat = 20 MPa√m is given. As Fig. 8c shows, the
resulting fatigue life increases significantly with higher values Kmat, even if the value of Kmat is just
a little higher (the step size for Kmat in Fig. 8c is 5 MPa√m). The other parameter values for the
calculation of the crack propagation and welded components (R > 0.5), according to [11], are (also
used in the calculations): ΔKth = 2 MPa√m, C = 1.37e-7, m = 2.25.

5 MULTI-LEVEL FRAMEWORK FOR FATIGUE LIFE PREDICTIONS

5.1 General remarks


A multi-level framework was developed in [1], to give guidance for engineers to evaluate the
service life of critical details of steel railway bridges. Its modular design enables the integration of
individual approaches, according to the level of knowledge about the real traffic loads, the material
parameters, or the load history. Different levels of detailing allow a differentiation, with regard to
the calculation effort and potential laboratory tests. The idea is, that a higher effort increases the
calculated fatigue life. Different levels in the individual steps are possible. It is also possible to
integrate future research findings in this framework and to estimate the fatigue life of components
with or without cracks. The results are the basis for inspection intervals, necessary reinforcements,
or a time horizon for a new construction.
5.2 Necessary preliminary studies
The critical details have to be defined, that means cross sections with the highest utilisation factor
of a conventional fatigue life assessment approach, or cross sections with detected fatigue cracks.
5.3 Step 1: Determination of the equivalent fatigue stress range Δse
· Level 1: Based on the Eurocodes load model UIC71 and the damage equivalent factor l in [2].
· Level 2: Based on the simulation of the crossing of the 8 fatigue train types of the Eurocode
[2], calculation of a total stress spectrum and the equivalent stress range Δse.
· Level 3: Based on real measured traffic load effects over a representative time (Δse based on
measured stress spectra).
Maybe a differentiation between different periods of the service life is necessary (past, present,
future, before or after reinforcements).
5.4 Step 2: Determination of the fatigue resistance
· Level 1: Based on the nominal stress concept, with a detail classification based on [4].
· Level 2: Based on the geometric or the effective notch stress concept.

5.5 Step 3: Results at the time of the investigation


Evaluation of the combined results of step 1 and 2, leading to a remaining fatigue life. Is it possible
to omit step 4 and 5, if a sufficient remaining fatigue life can be verified.
5.6 Step 4: Non-destructive testing of the critical details
If any cracks are found with non-destructive testing (NDT), the fatigue life can be determined with
fracture mechanics approaches (step 5). If no cracks are found, it is recommended, either to look for
© Ernst & Sohn Verlag für Architektur und technische Wissenschaften GmbH & Co. KG, Berlin ∙ CE/papers (2017)
| 2397

improved assumptions in step 1 to 2, or to assume an initial crack that has just not been found,
because of a crack size below the detection limit of the NDT method and continue to step 5.
5.7 Step 5: Application of fracture mechanics
a.) Determination of the necessary fracture mechanics material parameters.
· Level 1: Based on a literature values like in [11] or [12].
· Level 2: Based on specific fracture mechanics tests in a laboratory.
b.) Determination of the critical crack size with the FAD concept and the reduced combination of
actions, according to EN 1993-1-10 [14] and additional residual stresses for welded details of about
100 MPa (background of [14]).
c.) Remaining fatigue life, based on crack propagation procedure. Definition of the maximum stress
range and the shape of the stress spectrum, to calculate the fatigue life from the measured or defined
initial crack size to the critical crack size.
· Level 1: Application of constant amplitude loading, with an estimation of Δs with a simple
numerical model of the bridge and a conservative position of the load model UIC71.
· Level 2: Application of the simulated stress spectrum from step 1, level 2.
· Level 3: Application of the measured stress spectrum from step 1, level 3.
The effect of residual stresses is already included in the crack propagation parameters for R > 0.5.

REFERENCES
[1] H. Unterweger, P. Kugler, Wirtschaftliche Prognose der Restlebensdauer von stählernen
Eisenbahnbrücken mit offener Fahrbahn, Gesamtbericht des FFG Forschungsprojekts Nr. 845668,
Institut für Stahlbau, Technische Universität Graz, Oktober 2016
[2] EN 1991-2:2003-09 + AC:2010-02, Eurocode 1: Actions on structures - Part 2: Traffic loads on
bridges, European Committee for Standardization, Brussels, 2010
[3] EN 1993-2:2006-10 + AC:2009-07, Eurocode 3: Design of steel structures - Part 2: Steel Bridges,
European Committee for Standardization, Brussels, 2009
[4] EN 1993-1-9:2005-05 + AC:2009-04, Eurocode 3: Design of steel structures - Part 1-9: Fatigue,
European Committee for Standardization, Brussels, 2009
[5] E. Haibach, Betriebsfestigkeit - Verfahren und Daten zur Bauteilberechnung, Springer Verlag, 2006
[6] M. Olschnegger, Verbesserungen beim Nachweis der Restlebensdauer von Eisenbahnbrücken,
Masterarbeit an der TU Graz, September 2015
[7] D. Radaj, C. M. Sonsino, W. Fricke, Fatigue assessment of welded joints by local approaches, Second
edition, Woodhead Publishing Ltd., Cambridge, 2006
[8] D. Radaj, M. Vormwald, Ermüdungsfestigkeit, 3. Auflage, Springer Verlag, 2007
[9] A. F. Hobbacher, Recommendations for Fatigue Design of Welded Joints and Components, Second
Edition, International Institute of Welding, Springer Verlag, 2016
[10] T. L. Anderson, Fracture Mechanics - Fundamentals and Applications, Third Edition, Taylor & Francis
Group, 2005
[11] BS 7910:2013 + A1:2015, Guide to methods for assessing the acceptability of flaws in metallic
structures, The British Standards Institution, 2015
[12] FKM Richtlinie, Bruchmechanischer Festigkeitsnachweis, VDMA Verlag, 2009
[13] B. Kühn, Beitrag zur Vereinheitlichung der europäischen Regelungen zur Vermeidung von Sprödbruch,
Shaker Verlag, Aachen, 2005
[14] EN 1993-1-10:2005-05 + AC:2009-03, Eurocode 3: Design of steel structures - Part 1-10: Material
toughness and through-thickness properties, European Committee for Standardization, Brussels, 2009

© Ernst & Sohn Verlag für Architektur und technische Wissenschaften GmbH & Co. KG, Berlin ∙ CE/papers (2017)

You might also like